14:00:11 <shardy> #startmeeting tripleo
14:00:17 <openstack> Meeting started Tue May  3 14:00:11 2016 UTC and is due to finish in 60 minutes.  The chair is shardy. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:18 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:21 <openstack> The meeting name has been set to 'tripleo'
14:00:28 <shardy> #topic rollcall
14:00:30 <jrist> o/
14:00:33 <shardy> Hi all, who's around?
14:00:36 <shadower> hey
14:00:36 <sshnaidm> o/
14:00:39 <rhallisey> hi
14:00:41 <trown> o/
14:00:44 <akrivoka> hello \o
14:00:56 <derekh> o/
14:01:09 <slagle> hi
14:01:14 <EmilienM> o/
14:01:44 <pabelanger> o/
14:01:46 <dprince> hi
14:02:14 <shardy> Ok, lets get started :)
14:02:27 <shardy> #topic agenda
14:02:40 <shardy> #link https://wiki.openstack.org/wiki/Meetings/TripleO
14:02:51 <shardy> Anyone have any one-off items to add?
14:03:04 <shardy> I added one re summit sessions, and trown has one re puppet packages
14:03:19 <trown> shardy: I think sshnaidm added one about tempest as well
14:03:25 <jrist> shardy: can you just go over the status of the subteams? - I don't recall a final decision
14:03:28 * shardy refreshes ;)
14:03:28 <sshnaidm> shardy, trown yes, right
14:04:09 <shardy> jrist: I don't think there has been one, other than to say it's fine to organize temporary teams around features
14:04:09 * beagles kicks IRC client
14:04:14 <beagles> o/
14:04:18 <jrist> shardy: sounds good. thanks
14:04:20 <shardy> such as EmilienM has started for the composable roles work
14:04:26 <shardy> (see openstack-dev thread)
14:04:51 <shardy> #topic one off agenda items
14:05:08 <shardy> #info (trown)Puppet modules from source or packages? (packages are now individual modules and work like every other dlrn package)
14:05:17 <trown> #link https://trunk.rdoproject.org/centos7-master/report.html
14:05:32 <shardy> trown: so we can switch over from the source elements now?
14:05:36 <trown> we have puppet modules packaged individually now in our master trunk repo in RDO
14:05:38 <slagle> trown: does that mean we can default to packages now?
14:05:48 <trown> shardy: slagle that is my question :)
14:05:50 <slagle> or that's what you're asking :)
14:05:52 <shardy> trown: will this only be for master/Newton ?
14:05:54 <slagle> heh, yea, just realized that
14:06:05 <trown> shardy: ya and moving forward
14:06:12 <shardy> +1 for switching, should save us some time building stuff in CI
14:06:19 <slagle> i'd prefer to default to packages if we can, so it's consistent
14:06:23 <shardy> we do still have the o-p-m issue for stable branches tho
14:06:32 <trown> I have a patch up for tripleo-quickstart to use puppet from source, but it is a bit obsolete by this
14:06:43 <shardy> I proposed one fix for that, but it needs revisiting as there was a preference for a different approach
14:07:12 <dprince> trown: where are the package sources?
14:07:14 <dprince> trown: https://github.com/openstack-packages/
14:08:17 <derekh> shardy: it mightn't save us time building puppet modules, as we don't currently build any but +1 for consistency
14:08:22 <trown> dprince: hmm good question... everything in RDO moved to review.rdoproject.org and https://github.com/rdo-packages/ but I dont see puppet there
14:08:46 <shardy> derekh: Yeah I meant all the dib stuff we do around modules
14:08:54 <dprince> trown: before moving I would like to know where we push fixes, etc. given it would block us upstream potentially
14:08:57 <derekh> shardy: ack
14:09:16 <dprince> trown: also, while I like the consistency I would sort of like the puppet modules to pave the way for us here
14:09:32 <dprince> trown: in other words... do we plan on doing the same for the puppet-* CI jobs as well?
14:09:33 <slagle> yea, before we switch, we definitely want to prove we're testing proposed patches to the puppet modules
14:09:43 <slagle> (building the test package via dlrn, etc)
14:10:04 <shardy> Ok, should we follow up on this with some ML discussion, it sounds like we need to break it down into a few steps
14:10:07 <shardy> ?
14:10:10 <trown> +1
14:10:16 <dprince> shardy: yep, thanks
14:10:29 <trown> just wanted to get some feedback on if the direction made sense
14:10:44 <trown> or if there was some hard "we must use puppet from source"
14:10:55 <shardy> #info direction to use puppet module packages agreed, more discussion re how to switch on ML pending
14:11:17 <shardy> trown: the main hard requirement is to keep Depends-On: $puppet-module-change working
14:11:38 <shardy> most of that should just work, but there's going to be some hacking on tripleo.sh at least
14:11:40 <trown> shardy: ya, and if we can build with dlrn that should be fine
14:11:58 <shardy> #info (shardy)Summit session retrospective
14:12:13 <shardy> So, this item was just a request for feedback on how folks thought the sessions went
14:12:21 <shardy> in particular the number and type of sessions
14:12:31 <shardy> (to help when we're asked what we want next summit)
14:13:10 <shardy> Do folks feel we had about the right number of sessions this time, or too much/little?
14:13:33 <slagle> i think it was about right imo
14:13:34 <shardy> I felt it was about right myself - some projects had a lot more, but we covered most of the major topics
14:13:36 <dprince> shardy: I was happy with it. we could probably do with one more session though
14:13:48 <slagle> our fishbwols and working sessions felt like the same format though
14:14:06 <dprince> slagle: fishbowls get you a bigger room though :)
14:14:07 <slagle> so perhaps making our working sessions more about "doing work" next time
14:14:16 <slagle> (if we find that valuable)
14:14:19 <shardy> Yeah, I thought the same thing - all of them were open discussion, which is good, but I'm not sure the fishbowls really encouraged participation
14:14:34 <EmilienM> next time I'll give you sessions from puppet, we did not used them all
14:14:58 <dprince> + for more work sessions
14:15:19 <bandini> +1 for more work sessions, yes
14:15:26 <slagle> i guess our fishbowls felt kind of presentation-y and our work sessions were more like the traditional fishbowl
14:15:32 <shardy> #info possbly more work sessions next summit, and consider a more granular agenda to enable "doing work" vs open discussion
14:15:35 <gfidente> I think it'd be nice for fishbowls to come up with actual questions in the etherpad
14:15:38 <slagle> so however we address that :)
14:15:45 <gfidente> to which we try to find an answer before time runs out
14:15:53 <shardy> slagle: I felt the same, but it was mostly due to the size of the rooms vs the content I think?
14:16:05 <derekh> I wouldn't be opposed to tripleo getting it own room Tuesday/Wednesday to do some side work sessions while we are all in the same place
14:16:20 <slagle> shardy: could be yea
14:16:29 <shardy> #info investigate dedicated space for pre-meetup meetup
14:16:38 <d0ugal> o/
14:16:48 <shardy> Ok, that's all good feedback, thanks - anything else before we move on?
14:17:11 <gfidente> shardy, specific questions in the pads, not topic too generic
14:17:24 <gfidente> or we end up with more confusion and not many action items
14:17:41 <gfidente> I noticed you were indeed trying to sum up the action items yourself in multiple occasions
14:17:43 <shardy> #info attempt to get more focussed questions ahead of sessions, and clear topic definition
14:18:26 <shardy> gfidente: Yeah, I was happy to do that though, assuming folks found it useful
14:18:40 <shardy> Ok sshnaidm , you're up next:
14:18:45 <shardy> #info (sshnaidm) tempest passed https://review.openstack.org/#/c/295844/ ... is tripleo-ci where the py code to generate tripleo.conf belongs to?
14:18:58 <sshnaidm> shardy, yes, thanks
14:19:35 <sshnaidm> My position is let's start from Ci to see hot it goes, then let's think how to do it in tripleo. What is good for CI is not always good for customer/developer, maybe we'll use completely different tools to configure tempest. So let's keep it in CI repo for now
14:19:47 <sshnaidm> gfidente, ?
14:19:51 <shardy> So, the issue here is that there's not one common way to configure tempest, right?
14:19:59 <sshnaidm> shardy, yes
14:20:01 <gfidente> yeah and we seem to want tripleoclient to do that
14:20:13 <gfidente> because it already does dump some stuff, but that's clearly not enough
14:20:20 <shardy> Can anyone explain why there isn't e.g puppet support for it?
14:20:33 <gfidente> so my point is that we *could* move that code in the tripleoclient
14:20:40 <shardy> I personally would prefer we didn't hack all this into tripleoclient
14:20:41 <gfidente> (or even tripleo-common)
14:20:46 <dprince> shardy: that would imply we wanted it installed on the undercloud (always)
14:20:48 <gfidente> yeah me as well
14:20:53 <EmilienM> shardy: upstream puppet CI uses puppet-tempest.
14:20:56 <dprince> shardy: I don't think of tempest as an undercloud only tool
14:20:58 <shardy> dprince: well, you can run puppet elsewhere
14:21:01 <sshnaidm> shardy, I wrote about puppet in openstack-dev, it's not good solution
14:21:20 <dprince> shardy: sure, but we aren't providing a way to do that
14:21:29 <shardy> dprince: do we have to though?
14:21:37 <trown> EmilienM: that is not 100% accurate... upstream puppet CI uses puppet-tempest and a bunch of other steps to set it up
14:21:43 <shardy> like, for CI purposes, and many operators, configuring it on the undercloud is OK
14:21:45 * dprince doesn't like the enable_tempest option in instack
14:22:06 <EmilienM> trown: puppet-tempest has in charge of tempest.conf management
14:22:09 <shardy> Folks are still free to configure it $somewhere_else via whatever method they like
14:22:19 <EmilienM> right ^
14:22:54 <trown> EmilienM: there is a very thorough exploration of using puppet-tempest on the ML by sshnaidm
14:23:08 <dprince> shardy: for the same reasons I didn't want people going and adding Ansible dynamic discovery to the undercloud. Because once you put it there everyone just shells into the undercloud for everything
14:23:12 <trown> it is not so simple to use
14:23:20 <EmilienM> trown: cool, so let's not use it, I don't mind;
14:23:36 <gfidente> :(
14:23:52 <shardy> dprince: Ok, so perhaps we shouldn't configure tempest at all, and just implement a solution inside tripleo-ci?
14:24:14 <sshnaidm> let me please to summarize a little, there is a patch with configure_tempest.py in CI repo which configure tempest and run tests: https://review.openstack.org/#/c/295844/  , please look at it in your time, try it
14:24:18 <gfidente> that is what the existing submission dows fwiw, and it works well as well
14:24:22 <dprince> shardy: that, or I'd prefer a more "client" based solution. Which makes it more clear that you can install it anywhere...
14:24:30 <trown> I am +1 to tripleo-ci in the short term, with the hope that tempest figures this out in the longer term
14:25:15 <shardy> dprince: Ok, so you advocate the python client also acting as a local config generation tool
14:25:28 <dprince> shardy: I think that would make sense
14:25:43 * derekh is also happy to go with a tripleo-ci solution in the short term
14:25:53 <dprince> shardy: I'd almost say this sounds like a missing part of tempest
14:25:55 <EmilienM> trown: I'm not sure it will happen, honestly
14:25:59 <dprince> shardy: like a python-tempestclient or something
14:26:07 <EmilienM> trown: this topic is on the table for very long time
14:26:09 <shardy> personally I see this as pretty out-of-scope for python-tripleoclient, but it seems like a pretty non-tripleo-specific thing to have in our tripleo-centric client
14:26:21 <dprince> shardy: I'd prefer a more animated name... but at least that would tell people what it did
14:26:23 <trown> EmilienM: ya, but that is clearly where it should be...
14:27:06 <shardy> Ok, we may have to timebox this topic a little - can we vote on anything concrete here today, or does this need to continue on the ML?
14:27:43 <sshnaidm> let's vote? :)
14:27:43 <trown> I think there is general positivity towards sshnaidm patch to tripleo-ci in the short term
14:28:18 <shardy> sshnaidm: vote on what exactly - that we're OK with adding this into tripleo-ci as a first step?
14:28:27 <sshnaidm> shardy, yep
14:28:54 <shardy> I'm +1 on that, as I think the *client discussion is much harder, and we'll commit to interfaces if we do that
14:30:01 <shardy> Ok, lets move on
14:30:05 <derekh> +1 also
14:30:11 <dprince> tripleo-ci is a great first step
14:30:15 <shardy> #topic bugs
14:30:40 <shardy> So I sent a ML message that I've started trying to clean up the bug backlog
14:31:24 <shardy> I marked a lot of things incomplete which had been open a really long time without activity - but there's still a ton of old stuff in there we need to purge
14:32:02 <shardy> Any help purging the remaining old and now not relevant stuff would be much appreciated
14:32:51 <shardy> #link https://bugs.launchpad.net/tripleo
14:33:00 <shardy> Anything else bug related folks want to raise?
14:33:06 <sshnaidm> shardy, for this you need the permissions, right?
14:33:32 <shardy> sshnaidm: I created a tripleo-drivers group, which is currently only tripleo-core
14:33:41 <derekh> shardy: so you want to to look through bugs a close what loks irrelevant?
14:33:42 <sshnaidm> shardy, ok
14:33:59 <shardy> If non core folks want to help with bug triage that's fine, I can add you
14:34:25 <shardy> derekh: Yes, I'm trying to purge really old things not relevant to our current implementation
14:34:27 <sshnaidm> shardy, I'd be happy
14:34:42 <shardy> we had like 400 open bugs, and trying to targed what was relevent to Newton is basically impossible
14:34:45 * derekh will help out
14:34:57 <hewbrocca> There's something to be said for "close them all, if they're real someone will re-open them"
14:35:07 <hewbrocca> Maybe that's a bit extreme...
14:35:11 <beagles> I'll pick at them as well...
14:35:20 <shardy> hewbrocca: That's kind of what I did, with a script that set everything pre-liberty incomplete with a message
14:35:24 <beagles> we don't require permissions to mark as incomplete though do we?
14:35:26 <EmilienM> shardy: we could organize a sprint of triage
14:35:31 <EmilienM> I did that with puppet team during summit
14:35:31 <derekh> shardy: I closed a few the other day, its more time consuming then I thought but worth it
14:35:40 <shardy> beagles: No, you don't, or even better invalid ;)
14:35:43 <EmilienM> we splited the list of bugs, and did triage
14:35:49 <beagles> shardy: ack
14:36:08 <EmilienM> we also had nearly 300 bugs
14:36:27 <EmilienM> now, we have less than 50
14:37:01 <shardy> EmilienM: Yeah, that's the sort of progress I'm hoping we can make - perhaps an organized bug day would be good, but lets try with a few folks helping out for now
14:37:15 <gfidente> wasn't the bot posting about untriaged bugs at some point? would that help spread the triaging over time and avoid this situation in the future?
14:37:42 <shardy> gfidente: It's not untriaged bugs, it's stuff that's been triaged then sat there for 2 years
14:38:05 <gfidente> oh from the non-puppet implementation as well then?
14:38:05 <shardy> Also, I created newton milestones, so if folks can assign them to newton when triaging, that would be good:
14:38:08 <shardy> https://launchpad.net/tripleo/+milestone/newton-1
14:38:22 <shardy> that will make it easier to decide when is a good time to cut interim/milestone releases
14:38:29 <gfidente> shardy yeah totally +1 on that
14:38:30 <shardy> gfidente: Yes, exactly
14:38:42 <shardy> Ok, unless there's anything else on bugs, shall we move on?
14:39:06 <shardy> #topic Projects releases or stable backports
14:39:47 <shardy> So, slagle and I asked for some changes to the release tagging:
14:39:50 <shardy> https://review.openstack.org/#/c/308236/
14:40:00 <shardy> https://review.openstack.org/#/c/308574
14:40:14 <shardy> basically we'll be cycle-trailing to match what the puppet team is doing
14:40:31 <shardy> and also assert that we follow stable branch policy, with interim releases
14:40:55 <shardy> Feel free to add feedback on those if you'd like
14:41:10 <shardy> Anything else release related to discuss?
14:42:31 <shardy> #topic CI
14:42:52 <shardy> derekh: did you want to give a summary of the plans wrt the CI racks etc?
14:42:52 <derekh> We got bitten by new dependencies for paramiko on friday so CI was out for a few days, its mostly fixed now, but I think nodepool still wont be able to build is nightly F22 image, so that still needs looking at I think
14:43:04 <derekh> shardy: sure can do
14:43:37 <derekh> Basically the current plan is to move to RDO cloud before the rack move,
14:44:01 <derekh> then let RDO cloud absorb our HW
14:44:25 <derekh> I'm not sure yet exactly how much quota we'll get before they take over our HW
14:44:35 <shardy> derekh: do we have any data re how that's likely to impact our recent performance/capacity issues?
14:44:47 <derekh> but we'll probably have to reduce are usage for a time
14:45:01 <shardy> Yeah, that's what I was worried about ;)
14:45:07 <derekh> performance - don't know
14:45:28 <shardy> We've had several discussions around optimizing for performance, e.g more memory, SSDs etc
14:45:35 <derekh> capacity - reduced for a time, untill they redeploy our HW
14:45:37 <hewbrocca> My response on this FWIW was
14:45:47 <hewbrocca> leave the rack in place until RDO-cloud is really ready
14:46:01 <shardy> it'd be good to ensure we'll at least maintain comparable CI runtimes on the new cloud
14:46:07 <pabelanger> I am also working on tripleo-centos-7 replacement slave for tripleo-fedora-22: https://review.openstack.org/#/c/311721/
14:46:27 <pabelanger> image is uploaded to tripleo cloud, but failing to boot ATM (missing AFS mirror). Have a patch to fix that
14:46:40 <shardy> pabelanger: sounds good, thanks!
14:46:43 <pabelanger> https://review.openstack.org/#/c/312058/
14:46:57 <trown> derekh: do we have a date for the rack move?
14:46:59 <slagle> i also hope to investigate using multi-node jobs in nodepool directly. if anyone wants to help with this, lets sync up
14:47:13 <hewbrocca> trown: we've got options on the date
14:47:20 <hewbrocca> we can actually wait as long as next year
14:47:26 <derekh> hewbrocca: yup, just seeing your mail now, no need to rush the move but we have a Hard deadline
14:47:32 <derekh> trown: July 1st
14:47:50 <derekh> hewbrocca: ya, didn't know that?
14:47:52 <trown> oh... then definitely +100 to not moving before we have at least some jobs running in RDO-cloud
14:48:02 <derekh> hewbrocca: that changes things
14:48:48 <derekh> as far as RAM and SSD upgrades , we're waiting on orders to arrive, so that should be happening soon
14:50:03 <shardy> derekh: I guess my point was, if we do those upgrades, and the nodes are then added to RDO cloud, will we be able to get dedicated access to that pool of nodes
14:50:16 <hewbrocca> shardy: that would be the idea, yes
14:50:31 <shardy> hewbrocca: ack, sounds good then, thanks for clarifying
14:50:50 <shardy> Ok, anything else CI related before we move on?
14:51:10 <derekh> pabelanger: ack, thanks
14:51:16 <derekh> Periodic jobs failed last night has anybody looked into it? Once we get it passing we can start using a cached overcloud images.
14:51:40 <dprince> derekh: not yet, but I will have a look afterwards
14:51:54 <shardy> #topic Specs
14:52:08 <shardy> So, perhaps we should change that topic to just "features" or something
14:52:21 <pabelanger> dprince: derekh: likely related to static.o.o filesystem issue
14:52:42 <dprince> pabelanger: ack, will keep that in mind
14:52:48 <shardy> IMO we shouldn't enforce specs for all features this cycle, and we did agree a while back to try the spec-lite bug approach similar to how glance and other projects do it
14:52:53 <shardy> https://bugs.launchpad.net/tripleo/+bugs?field.tag=spec-lite
14:53:12 <shardy> I created the tag, so if you want to add a minor RFE, you can do it as a bug and tag it appropriately
14:53:28 <shardy> in particular it'd be good to capture a laundry-list of usability issues via this method
14:53:43 <slagle> if a spec-lite much different from a blueprint?
14:53:53 <slagle> *is
14:53:56 <shardy> slagle: Not really, other than it's easier to have a discussion in a bug
14:54:02 <shardy> https://blueprints.launchpad.net/tripleo
14:54:25 <shardy> blueprints are also still fine, and I've targetted most of those which weren't expired/obsolete to newton
14:54:53 <shardy> tbh I don't really mind what method folks use provided it's tracked somewhere, and targetted to Newton
14:54:57 <slagle> shardy: ok, i filed one i'd like to tackle in newton: https://blueprints.launchpad.net/tripleo/+spec/undercloud-elements
14:55:10 <shardy> then we can reasonably see what's on the roadmap, and figure out how we're doing ref the backlog
14:55:12 <slagle> hopefully that is uncontentious :)
14:55:42 <shardy> Ok, 5mins - anything else spec/feature related, or shall we go to open discussion?
14:56:01 <beagles> I added a bp to add dvr support last night
14:56:18 <shardy> Ok, I'll trage the new BPs today, thanks!
14:56:22 <shardy> triage even
14:56:42 <shardy> #topic Open Discussion
14:56:48 <beagles> but it's kind of segue into the open discussion
14:56:50 <shardy> Anything folks want to raise?
14:57:00 <dprince> one new tripleo.org thing
14:57:06 <dprince> So I added a blogs page here
14:57:10 <dprince> #link http://tripleo.org/blog-posts.html
14:57:18 <rhallisey> I haven't finished any container specs yet.  Will try and get them up soon
14:57:23 <dprince> the idea being to highlight tripleo specific blogs
14:57:23 <beagles> I've been asked about the composable roles changes and time estimates etc.
14:57:24 <trown> slagle: for that instack-undercloud bp would we get rid of the puppet element eventually too? and just call puppet apply?
14:57:38 <trown> seems a bit weird to use DIB for just that one element
14:57:43 <shardy> dprince: Nice, I'm doing a session summary from Summit which I'll probably blog and post to openstack-dev
14:57:46 <beagles> dprince: nice
14:57:49 <slagle> trown: likely, yea
14:57:55 <dprince> need to iterate on figuring out how to better integrate it w/ planet.openstack.org but this is a start: https://review.openstack.org/#/c/311910/
14:57:56 <trown> cool
14:57:56 <slagle> trown: a bit further down the road probably
14:58:04 <slagle> trown: then instack-undercloud wouldnt use instack :)
14:58:12 <slagle> and we can declare success
14:58:23 <derekh> dprince: nice, should we come up with a list of potential blogging topics?
14:58:23 <trown> slagle: lol ya, that is what I was thinking... a bit odd :)
14:58:25 <dprince> shardy: nice, this is a way to crack the blogging whip a bit. On myself included
14:58:40 <dprince> derekh: yep, will do. Collaborative blogging man!
14:58:45 <shardy> beagles: see the mail from EmilienM, there's a group of folks working on composable services, and I'm looking into ways to do fully composable roles via some template pre-processing
14:59:39 <beagles> shardy: yup... generally I think we are going to need some idea of how to set expectations for new things like the DVR for newton
14:59:41 <slagle> dprince: the text boxes formatting is messed up
14:59:53 <shardy> beagles: OK, let's continue in #tripleo
14:59:57 <beagles> shardy: and other services as well of course
14:59:58 <dprince> slagle: sounds like a -1 to me :)
14:59:59 <beagles> yup
15:00:01 <shardy> Times up folks, thanks!
15:00:05 <shardy> #endmeeting